The MI Procedure
- Overview
- Getting Started
-
Syntax
-
DetailsDescriptive StatisticsEM Algorithm for Data with Missing ValuesStatistical Assumptions for Multiple ImputationMissing Data PatternsImputation MethodsMonotone Methods for Data Sets with Monotone Missing PatternsMonotone and FCS Regression MethodsMonotone and FCS Predictive Mean Matching MethodsMonotone and FCS Discriminant Function MethodsMonotone and FCS Logistic Regression MethodsMonotone Propensity Score MethodFCS Methods for Data Sets with Arbitrary Missing PatternsChecking Convergence in FCS MethodsMCMC Method for Arbitrary Missing Multivariate Normal DataProducing Monotone Missingness with the MCMC MethodMCMC Method SpecificationsChecking Convergence in MCMCInput Data SetsOutput Data SetsCombining Inferences from Multiply Imputed Data SetsMultiple Imputation EfficiencyNumber of ImputationsImputer’s Model Versus Analyst’s ModelParameter Simulation versus Multiple ImputationSensitivity Analysis for the MAR AssumptionMultiple Imputation with Pattern-Mixture ModelsSpecifying Sets of Observations for Imputation in Pattern-Mixture ModelsAdjusting Imputed Values in Pattern-Mixture ModelsSummary of Issues in Multiple ImputationPlot Options Superseded by ODS GraphicsODS Table NamesODS Graphics
-
ExamplesEM Algorithm for MLEMonotone Propensity Score MethodMonotone Regression MethodMonotone Logistic Regression Method for CLASS VariablesMonotone Discriminant Function Method for CLASS VariablesFCS Methods for Continuous VariablesFCS Method for CLASS VariablesFCS Method with Trace PlotMCMC MethodProducing Monotone Missingness with MCMCChecking Convergence in MCMCSaving and Using Parameters for MCMCTransforming to NormalityMultistage ImputationCreating Control-Based Pattern Imputation in Sensitivity AnalysisAdjusting Imputed Continuous Values in Sensitivity AnalysisAdjusting Imputed Classification Levels in Sensitivity AnalysisAdjusting Imputed Values with Parameters in a Data Set
- References
Input Data Sets
You can specify the input data set with missing values by using the DATA= option in the PROC MI statement. When an MCMC method is used, you can specify the data set that contains the reference distribution information for imputation with the INEST= option, the data set that contains initial parameter estimates for the MCMC method with the INITIAL=INPUT= option, and the data set that contains information for the prior distribution with the PRIOR=INPUT= option in the MCMC statement.
When the ADJUST option is specified in the MNAR statement, you can use the PARMS= option to specify the data set that contains adjustment parameters for the sensitivity analysis.
DATA=SAS-data-set
The input DATA= data set is an ordinary SAS data set that contains multivariate data with missing values.
INEST=SAS-data-set
The input INEST= data set is a TYPE=EST data set and contains a variable _Imputation_ to identify the imputation number. For each imputation, PROC MI reads the point estimate from the observations with _TYPE_=‘PARM’ or _TYPE_=‘PARMS’ and the associated covariances from the observations with _TYPE_=‘COV’ or _TYPE_=‘COVB’. These estimates are used as the reference distribution to impute values for observations in the DATA= data set. When
the input INEST= data set also contains observations with _TYPE_=‘SEED’, PROC MI reads the seed information for the random number generator from these observations. Otherwise, the SEED=
option provides the seed information.
INITIAL=INPUT=SAS-data-set
The input INITIAL=INPUT= data set is a TYPE=COV or CORR data set and provides initial parameter estimates for the MCMC method. The covariances derived from the TYPE=COV/CORR data set are divided by the number of observations to get the correct covariance matrix for the point estimate (sample mean).
If TYPE=COV, PROC MI reads the number of observations from the observations with _TYPE_=‘N’, the point estimate from the observations with _TYPE_=‘MEAN’, and the covariances from the observations with _TYPE_=‘COV’.
If TYPE=CORR, PROC MI reads the number of observations from the observations with _TYPE_=‘N’, the point estimate from the observations with _TYPE_=‘MEAN’, the correlations from the observations with _TYPE_=‘CORR’, and the standard deviations from the observations with _TYPE_=‘STD’.
PARMS= SAS-data-set
The input PARMS= data set is an ordinary SAS data set that contains adjustment parameters for imputed values of the specified imputed variables.
The PARMS= data set contains variables _Imputation_ for the imputation number, the SHIFT= or DELTA= variable for the shift parameter, and the SCALE= variable for the scale parameter.
Either the shift or scale variable must be included in the data set.
PRIOR=INPUT=SAS-data-set
The input PRIOR=INPUT= data set is a TYPE=COV data set that provides information for the prior distribution. You can use
the data set to specify a prior distribution for 
of the form
![\[ \bSigma \sim W^{-1} \left( \, d^{*}, \, d^{*}\mb{S}^{*} \right) \]](images/statug_mi0286.png)
where 
is the degrees of freedom. PROC MI reads the matrix 
from observations with _TYPE_=‘COV’ and reads 
from observations with _TYPE_=‘N’.
You can also use this data set to specify a prior distribution for 
of the form
![\[ \bmu \sim N \left( \, \bmu _{0} ,\, \frac{1}{n_{0}} \bSigma \right) \]](images/statug_mi0288.png)
PROC MI reads the mean vector 
from observations with _TYPE_=‘MEAN’ and reads 
from observations with _TYPE_=‘N_MEAN’. When there are no observations with _TYPE_=‘N_MEAN’, PROC MI reads from observations with _TYPE_=‘N’.